Market Analysis — Bank Statement Analysis (BSA) / Cash-Flow Underwriting

Living competitive doc. Last updated: 2026-06-23. Owner: Salil. For sales/marketing. Pairs with gtm-plan.md. Pricing for most players is opaque (enterprise “contact us”); figures are flagged where they are third-party estimates rather than published rate cards.

The category in one paragraph

“Bank Statement Analysis” (BSA) — also sold as cash-flow underwriting, financial document intelligence, or transaction enrichment — turns a borrower’s bank statement (PDF, scan, CSV, or Account-Aggregator feed) into structured, decision-ready signals: income, obligations, FOIR, bounces, counterparties, fraud/tamper flags, and a risk view. In India it is a core part of every digital-lending stack and is increasingly bought by NBFCs, DSAs, fintech lenders, and CA firms.

The market’s shape (the opening for us)

1. Enterprise-sales-gated, opaque pricing. Perfios, CRIF, Signzy, ScoreMe, Ocrolus, Heron, Inscribe, MX — almost all hide pricing behind “book a demo,” sell annual contracts with volume minimums, and bundle BSA inside a larger onboarding/fraud suite.
2. Transparent self-serve is rare. Only Precisa (₹100/account), Pro Analyser, and FinBox’s free trial approach self-serve in India; DocuClipper/Nanonets do it in pure OCR.
3. The long tail is underserved. Individual CA firms, DSAs, brokers, and small NBFCs can’t justify an enterprise contract or an integration project. Precisa and Pro Analyser are the only ones courting them — and not sharply.
4. Account Aggregator (AA) changes ingestion, not the need for analysis. AA is growing fast (~120M accounts linked by Dec 2024) but covers only ~38% of borrowers; joint accounts, cooperative banks/RRBs, scanned and historical statements still need PDF. And AA data still needs the same categorization/FOIR/fraud layer BSA tools sell. PDF-first analysis stays relevant for years.

Competitor tracker — India

Competitor tracker — Global (context + feature benchmarks)

What vendors compete on (the axes)

• Accuracy (loudest in OCR: “99%+”), often via human-in-the-loop.
• Fraud / tamper detection — the fastest-growing battleground; AI-generated document fraud reportedly up ~5x in 2025 (Inscribe). Bank statements are 59% of flagged fraud docs.
• Cash-flow analytics depth (Ocrolus, Heron) — signals, not just text.
• Coverage — bank/format breadth; AA-native ingestion.
• Speed / automation — “30 min → seconds.”

Nobody in this list competes on verifiable, reproducible output — deterministic money math, a balance-reconciliation gate (now a dual gate: running balance + credit/debit column totals), deterministic extraction, a deterministic auditable risk score, an adversarial test harness. And no self-serve player goes past analysis into decision support a human owns: a deterministic approve/refer/decline recommendation against the lender’s policy, a delegated-authority override queue, an append-only audit trail, and CSV/LOS export. Enterprise suites (Perfios) have decisioning, but gated and black-box. That’s an open lane (see gtm-plan.md).

Obsrv scorecard vs the field

Where we win: frictionless self-serve, transparent pricing, a trust story built on verifiable numbers (deterministic extraction + a dual reconciliation gate + deterministic scoring, tested against attacks), borrower-level consolidation with auditable inter-account transfer dedup, and — the newest, sharpest edge — self-serve decision support a human owns: a deterministic approve/refer/decline recommendation against the lender’s own policy, a delegated-authority override queue, an append-only audit trail, and one-click CSV export for the audit file / LOS. No self-serve competitor goes past analysis into the decision and its paper trail; the enterprise suites that do are gated and opaque. Where we lose today: no AA yet, no tamper detection yet, no enterprise suite/bundling, and per-page is costlier than Precisa’s per-account for heavy multi-month pulls (see the pricing note).

Pricing (decided 2026-06-23): credit packs, not a per-borrower bundle

Precisa charges ₹100/account for up to 12 statements; our ₹5/page makes a full multi-month pull (say 72 pages = ₹360) ~3.6× their headline. We considered a per-borrower bundle to match them and chose not to:

• Model: pure-consumption credit packs (₹500/₹1,000/₹2,000, +0/5/10% bonus, ₹5/credit, 12-mo expiry, non-refundable). Simpler, on-brand (the AI-software model), and it keeps all features ungated so the moat stays visible. Margins hold ~62–65% (see pricing.md).
• Why not a bundle: a flat per-borrower price exposes us to page-count variance (our cost is LLM-linked ~₹1.8/page; Precisa’s is templated-OCR ~₹0), and bulk-bonus credits already give heavy users a better effective rate (Pro ≈ ₹4.55/credit).
• Positioning, not price-matching: we sell borrower underwriting (consolidate + decide + audit), a higher-value category than per-account OCR — and one Obsrv case covers a multi-account borrower that costs 2× on Precisa. We don’t race them to the OCR floor.
• The deliberate trade-off: for a pure casual single-statement checker we’re cheapest; for the heaviest multi-month/per-account shoppers Precisa’s flat ₹100 still looks cheaper on paper — we win them on value + auto-recharge/larger packs (later), not headline price.

Sources

India: perfios.com, precisa.in/price-plan, finbox.in/products/bankconnect, scoreme.in, crif.in, signzy.com, proanalyser.in, CGAP & Sahamati (AA stats), digitap.ai (AA-vs-PDF). Global: ocrolus.com (+ Vendr est.), herondata.io, inscribe.ai (2026 Fraud Report), plaid.com/pricing, docuclipper.com, nanonets.com. Full URLs in research notes.